A Survey of Machine Learning Techniques for Intelligent Proctoring Systems

The rapid shift to online education, accelerated by global events such as the pandemic, has underscored the critical need for robust and reliable online examination systems. A significant challenge in this paradigm is ensuring academic integrity and preventing malpractice. Traditional human proctoring methods, while effective in physical settings, are often labor-intensive, costly, and less scalable for remote assessments. This literature review paper systematically examines the advancements in automated online proctoring systems, with a particular focus on the application of deep learning techniques for detecting various forms of malpractice. We delve into different components of these systems, including face detection, multiple person detection, face spoofing, and head pose estimation. The paper synthesizes methodologies, datasets, and performance metrics from recent research, highlighting the evolution of deep learning models such as Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), and their variants in addressing the complexities of real-time cheating detection. Furthermore, it discusses the challenges and limitations of current systems, such as the need for more diverse and annotated datasets, ethical considerations, and the constant evolution of cheating strategies. Finally, we propose future research directions aimed at developing more sophisticated, adaptive, and ethical online proctoring solutions that leverage the full potential of artificial intelligence.